GitHub - mcclanahoochie/digpufft: Automatically exported from code.google.com/p/digpufft

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
README		README
digpufft-p3dfft-2.4.patch		digpufft-p3dfft-2.4.patch
example-setup.sh		example-setup.sh

Repository files navigation

//////////////////////
// Chris McClanahan //
//////////////////////


Download and install P3DFFT-2.4


Make sure that at least compiles (run a sample program for good measure)


Apply the .patch


Modify configure.ac variables:
  CUDA_INSTALL_PATH
  CUDA_SDK_INSTALL_PATH
to be correct for your setup


Re-run ./configure with new arguments:
  --enable-cuda
  --enable-fftw
  --enable-single 
  --enable-stride1
  
(add those to the end of whatever you used before)
(For now, those arguments MUST be used to enable CUDA)
  
  
New Makefiles with '-D_CUDA' should have been created, enabling CUFFT acceleration


An example setup script (example-setup.sh) is provided. It shows how to compile and patch p3dfft. Notice that this script is designed specifically for the Keeneland cluster (http://keeneland.gatech.edu/) and will probably need to be tweaked (e.g., different FFTW path, different compiler options) for your own environment.


Enjoy!